Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese-language deep-research workflow that is transparent about heavy multi-agent use, with the main user risk being time and token cost rather than harmful behavior.
Install only if you intentionally want a long, costly deep-research workflow. Before using it, specify the research question, maximum agents, maximum duration, model budget, and preferred language; avoid invoking it for ordinary quick research tasks.
The metadata exposes this as user-invocable while the description does not warn that it may switch to a stronger model, run for 4-72 hours, and consume substantial tokens. That omission undermines informed consent and increases the chance that users trigger an expensive long-running operation unintentionally.
The trigger phrases are broad enough to match ordinary research requests, which can cause the high-cost deep-research workflow to activate unexpectedly. In this skill, accidental invocation is more dangerous because the documented behavior includes switching to a stronger model, spawning 10-50 agents, and running for hours, creating cost and resource-exhaustion risk without clear user intent.
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
## 📚 相关文件
### 技能文档
- `skills/deep-research-mode/SKILL.md` - 本技能文档
### 工具脚本
- `tools/deep_research_orchestrator.py` - 多智能体协调器
The natural-language instructions, examples, and interaction patterns are entirely in Chinese, with no indication that users may choose another language. Under the policy, forcing a specific language without opt-in is a natural-language policy concern unless the locale restriction is explicitly documented and justified.
No suspicious patterns detected.